Reconstruct 12-lead ECG from II-lead： based on an improved generative adversarial network model

Electrocardiogram (ECG) provides essential clues for detecting heart diseases, and more and more smart devices are developed to monitor the ECG signals. However, due to the convenience of use, only limited leads of ECG signals are measured from such smart devices, which may significantly affect the validity of disease judgement. To enhance the diagnose performance of these smart devices, this study suggests a Generative Adversarial Network (GAN)-based approach that combines Transformer and U-Net structures to autonomously reconstruct the whole 12-lead ECG signals from a single lead ECG signals. This study evaluates the proposed model on the PTB-XL and the Shaoxing People's Hospital 12-lead ECG dataset, then compares it with several state of art approaches. The code for this study can be found in https://github.com/Chaoquan-123/12-lead-ECG-reconstruction.